Sure, but if the baseline level of annual usage is only 20% in the general population then you still have a significant correlation.

Correlation, and about a billion compounding factors.

For example: younger people are more likely to be regular cannabis users - and are also more likely than the general public to get in motor vehicle accidents.